Classifications

• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
  • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
  • G16H50/80—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for detecting, monitoring or modelling epidemics or pandemics, e.g. flu
• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
  • G16H10/40—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for data related to laboratory analysis, e.g. patient specimen analysis
• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
  • G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
  • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
• • G—PHYSICS
  • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
  • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
  • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
  • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N20/00—Machine learning
  • G06N20/20—Ensemble learning
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N5/00—Computing arrangements using knowledge-based models
  • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound

Definitions

• Prediction of disease is important not just in ensuring a timely treatment to the patients, but also in managing resources and logistics involved in the disease management process.
• the pandemic arising due to SARS-CoV-2 virus increased the burden on healthcare systems across the world.
• An accelerated growth in the number of individuals required to undergo testing for detection of SARS-CoV-2 virus resulted in shortage of infectious disease test kits such as reverse transcriptase polymerase chain reaction (RT- PCR) test kits, delay in providing test results to individuals, and delay in providing timely medical support to patients when presence of SARS-CoV-2 virus was not suspected.
• infectious disease test kits such as reverse transcriptase polymerase chain reaction (RT- PCR) test kits
• Prediction of presence of one or more pathogens such as a virus, e.g., SARS-CoV-2, bacteria, or the like in patients can be important in such scenarios.
• a pathogen such as a virus, e.g., SARS-CoV-2 virus in a subject
• a system for predicting a presence of a disease in a subject is disclosed.
• the system includes a processing unit and a memory. Additionally, the memory includes a disease prediction module configured to receive a plurality of parameters associated with the subject and determine a threshold for sensitivity and/or specificity associated with a trained machine learning model. Additionally, the module is configured to predict using the trained machine learning model the presence of the disease in the subject and the plurality of parameters associated with the subject and output the prediction on an output unit.
• a method for predicting a presence of a disease in a subject is disclosed. The method includes receiving a plurality of parameters associated with the subject. Additionally, the method includes determining a threshold for sensitivity and/or specificity associated with a trained machine learning model.
• the method includes predicting using a trained machine learning model the presence of the disease in the subject based on the sensitivity and specificity of the disease and the plurality of parameters associated with the subject. Further, the method includes outputting the prediction on an output unit.
• the present disclosure includes an article of manufacture including a non-transitory computer-readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method comprising: predicting a presence of a disease in accordance with the present disclosure.
• the present disclosure includes non-transient computer readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method comprising: predicting a presence of a disease in accordance with the present disclosure.
• the present disclosure includes predicting a presence of a disease in a subject in accordance with the present disclosure and subsequently treating the subject in need thereof.
• the present disclosure includes a system for predicting a presence of a disease in a population.
• the system includes a processing unit and a memory.
• the memory includes a disease prediction module configured to receive a plurality of parameters associated with one or more subjects and determine a threshold for sensitivity and/or specificity associated with a trained machine learning model. Additionally, the module is configured to predict using the trained machine learning model the presence of the disease in one or more subjects or a population of subjects and the plurality of parameters associated with the subject or population of subjects and output the prediction on an output unit.
• FIG.1 illustrates a system for predicting presence of a disease in a subject, according to an embodiment.
• FIG. 2 illustrates a process flow for predicting presence of a disease in a subject, according to an embodiment of the present disclosure.
• FIG.3 illustrates a receiver operating characteristic (ROC) curve for choosing threshold for sensitivity and specificity associated with a trained machine learning model for a requirement scenario, according to an embodiment of the present disclosure.
• FIG. 4 depicts a ROC curve of a 48-parameter ensemble model computed using a test set of the present disclosure.
• FIG. 5 depicts model interpretability: features arranged in decreasing order of importance as computed by the MIMIC LightGBM explainer.
• FIGS.6A, 6B and 6C depict use case embodiments such as in a pandemic scenario, and a use case for an endemic scenario at different regions (1, 2 and 3) of the ROC curve.
• Disclosed embodiments provide a system and method for predicting disease in a subject.
• Advantages of embodiments of the present disclosure include excellent and robust prediction of disease resulting from one or more pathogens to ensure, among other things, timely treatment to the patients or subjects in need thereof, a n d improved management of resources and logistics involved in the disease management process.
• Embodiments of the present disclosure reduce the burden on healthcare systems across the world, prevent test kit shortages, and shorten the duration on waiting for individual test results, and shorten the waiting period for medical support to patients when presence of a pathogen is not suspected.
• COVID-19 testing using RT-PCR (reverse transcriptase-polymerase chain reaction) and antigen tests scaled up massively since the onset of the COVID-19 pandemic, however, there are multiple scenarios where a CBC/Diff based predictive algorithm could be applied in the clinical value chain both in the pandemic and endemic scenarios.
• a CBC/Diff based predictive algorithm could act as a substitute for testing in countries dealing with unavailability or shortages of such tests.
• the same ML algorithm such as Algorithm A with different classification thresholds (such as a threshold value e.g., 0.4 to 1.0, 0.4-0.97, 0.4 to 0.95, 0.4-0.7, 0.4-0.6, 0.4-0.5, or 0.4- 0.45, 0.80-0.95, 0.99-1.0) can be used in multiple scenarios described herein.
• classification thresholds such as a threshold value e.g., 0.4 to 1.0, 0.4-0.97, 0.4 to 0.95, 0.4-0.7, 0.4-0.6, 0.4-0.5, or 0.4- 0.45, 0.80-0.95, 0.99-1.0
• the term “a” may be understood to mean “at least one”; (ii) the terms “comprising” and “including” may be understood to encompass itemized components or steps whether presented by themselves or together with one or more additional components or steps; and (iii) where ranges are provided, endpoints are included. [0023]
• the use of the term “at least one” will be understood to include one as well as any quantity more than one, including but not limited to, 2, 3, 4, 5, 10, 15, 20, 30, 40, 50, 100, etc.
• the term “at least one” may extend up to 100 or 1000 or more, depending on the term to which it is attached; in addition, the quantities of 100/1000 are not to be considered limiting, as higher limits may also produce satisfactory results.
• any reference to “one embodiment,” “an embodiment,” “some embodiments,” “one example,” “for example,” or “an example” means that a particular element, feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment.
• the appearance of the phrase “in some embodiments” or “one example” in various places in the specification is not necessarily all referring to the same embodiment, for example. Further, all references to one or more embodiments or examples are to be construed as non-limiting to the claims.
• the term “about” is used to indicate that a value includes the inherent variation of error for a composition/apparatus/ device, the method being employed to determine the value, or the variation that exists among the study subjects.
• the designated value may vary by plus or minus twenty percent, or fifteen percent, or twelve percent, or eleven percent, or ten percent, or nine percent, or eight percent, or seven percent, or six percent, or five percent, or four percent, or three percent, or two percent, or one percent from the specified value, as such variations are appropriate to perform the disclosed methods and understood by persons having ordinary skill in the art.
• A, B, C, or combinations thereof is intended to include at least one of: A, B, C, AB, AC, BC, or ABC, and if order is important in a particular context, also BA, CA, CB, CBA, BCA, ACB, BAC, or CAB.
• expressly included are combinations that contain repeats of one or more item or term, such as BB, AAA, AAB, BBC, AAABCCCC, CBBAAA, CABABB, and so forth.
• BB BB
• AAA AAA
• AAB BBC
• AAABCCCCCC CBBAAA
• CABABB CABABB
• CBC/Diff blood cell count with differential refers to a measure of the number of red blood cells, white blood cells, and platelets in the blood, including the different types of white blood cells (neutrophils, lymphocytes, monocytes, basophils, and eosinophils).
• white blood cells neurotrophils, lymphocytes, monocytes, basophils, and eosinophils.
• the amount of hemoglobin (substance in the blood that carries oxygen) and the hematocrit are also measured.
• diagnosis test is a step or series of steps that is or has been performed to attain information that is useful in determining whether a patient has a disease, disorder or condition and/or in classifying a disease, disorder or condition into a phenotypic category or any category having significance with regard to prognosis of a disease, disorder or condition, or likely response to treatment (either treatment in general or any particular treatment) of a disease, disorder or condition.
• diagnosis refers to providing any type of diagnostic information, including, but not limited to, whether a subject is likely to have or develop a disease, disorder or condition, state, staging or characteristic of a disease, disorder or condition as manifested in the subject, information related to the nature or classification of a tumor, information related to prognosis and/or information useful in selecting an appropriate treatment or additional diagnostic testing.
• Selection of treatment may include the choice of a particular therapeutic agent or other treatment modality such as surgery, radiation, etc., a choice about whether to withhold or deliver therapy, a choice relating to dosing regimen (e.g., frequency or level of one or more doses of a particular therapeutic agent or combination of therapeutic agents), etc.
• Disease and “disorder” are used interchangeably to refer to a condition in a subject including a harmful deviation from the normal structural or functional state of an organism.
• diseases/disorders include a subject having one or more viral infections, one or more bacterial infections, one or more fungal or parasitic infections, or sepsis.
• viral infection means the invasion by, multiplication and/or presence of a virus in a cell or a subject.
• a viral infection is an “active” infection, i.e., one in which the virus is replicating in a cell or a subject.
• Such an infection is characterized by the spread of the virus to other cells, tissues, and/or organs, from the cells, tissues, and/or organs initially infected by the virus.
• An infection may also be a latent infection, i.e., one in which the virus is not replicating.
• the term “bacterial infection” means the invasion by, multiplication and/or presence of a bacteria in a cell or a subject.
• the term “fungal infection” means the invasion by, multiplication and/or presence of a fungus in a cell or a subject.
• a biological sample refers to a biological sample obtained or derived from a human subject, as described herein.
• a biological sample includes biological tissue or fluid.
• a biological sample may include blood; blood cells; tissue or fine needle biopsy samples; cell-containing body fluids; free floating nucleic acids; cerebrospinal fluid; lymph; tissue biopsy specimens; surgical specimens; other body fluids, secretions, and/or excretions; and/or cells therefrom.
• a biological sample includes cells obtained from an individual, e.g., from a human or animal subject. In some embodiments, obtained cells are or include cells from an individual from whom the sample is obtained.
• a sample is a “primary sample” obtained directly from a source of interest by any appropriate means. For example, in some embodiments, a primary biological sample is obtained by methods selected from the group consisting of biopsy (e.g., fine needle aspiration or tissue biopsy), surgery, collection of body fluid (e.g., blood). In some embodiments, a sample is cardiac tissue obtained from the subject.
• sample refers to a preparation that is obtained by processing (e.g., by removing one or more components of and/or by adding one or more agents to) a primary sample. For example, filtering using a semi-permeable membrane.
• the sample may be a plasma sample that is treated with an anticoagulant selected from the group consisting of EDTA, heparin, and citrate.
• the sample may be processed to isolate one or more proteins (e.g., by capturing proteins with one or more antibodies).
• a “processed sample” may include, for example, nucleic acids or polypeptides extracted from a sample or obtained by subjecting a primary sample to techniques such as amplification or reverse transcription of mRNA, isolation and/or purification of certain components.
• SARS CoV-2 variant or “viral variant” or simply “variant” when used in reference to SARS CoV-2 (as will be apparent from context), is a SARS CoV-2 virus that includes one or more genetic changes relative to a SARS CoV-2 reference strain or one or more predominant viral variants already circulating in a population.
• a SARS CoV-2 variant includes one or more genomic mutations in an S gene sequence.
• a SARS CoV-2 variant includes one or more genomic mutations relative to a reference genome sequence or portion thereof. In some embodiments, a SARS CoV-2 variant includes one or more genomic mutations in an S gene sequence relative to an S gene sequence of a reference genome. In some embodiments, a reference genome sequence corresponds to that of the Wuhan-Hu1 strain (the first genetic sequence identified) or USA-WA1/2020 strain (the first identified in the United States) or portion thereof. In some embodiments, a SARS CoV-2 variant includes one or more genomic mutations relative to one or more predominant viral variants circulating in a population.
• a SARS CoV-2 variant includes one or more genomic mutations in an S gene sequence relative to an S gene sequence of a predominant viral variant circulating in a population.
• the term “subject” refers to an organism, for example, a mammal (e.g., a human). In some embodiments a human subject is an adult, adolescent, or pediatric subject. In some embodiments, a subject is at least 50, at least 55, at least 60, at least 65, at least 70, at least 75, or at least 80 years of age. In some embodiments, a subject is suffering from a disease, disorder or condition, e.g., a disease, disorder or condition that can be treated as provided herein.
• a subject is susceptible to a disease, disorder, or condition; in some embodiments, a susceptible subject is predisposed to and/or shows an increased risk (as compared to the average risk observed in a reference subject or population) of developing the disease, disorder or condition.
• a subject displays one or more symptoms of a disease, disorder or condition.
• a subject does not display a particular symptom (e.g., clinical manifestation of disease) or characteristic of a disease, disorder, or condition.
• a subject does not display any symptom or characteristic of a disease, disorder, or condition.
• a subject is a patient.
• a subject is an individual to whom diagnosis and/or therapy is and/or has been administered.
• a human subject is an adult, adolescent, or pediatric subject.
• a subject is at risk for a SARS CoV-2 viral infection.
• a subject is at risk for a viral infection, a bacterial infection, sepsis, or combinations thereof.
• a subject has been exposed or is suspected to have a COVID-19 infection.
• a subject is susceptible to COVID-19; in some embodiments, a susceptible subject is predisposed to and/or shows an increased risk (as compared to the average risk observed in a reference subject or population) of developing COVID- 19.
• a subject displays one or more symptoms of a SARS CoV- 2 infection. In some embodiments, a subject does not display a particular symptom or characteristic of COVID-19. In some embodiments, a subject does not display any symptom or characteristic of COVID-19 (i.e., is asymptomatic). In some embodiments, a subject is a patient. In some embodiments, a subject is an individual to whom diagnosis and/or therapy is and/or has been administered. [0038] As used herein, the term “substantially” means that the subsequently described event or circumstance completely occurs or that the subsequently described event or circumstance occurs to a great extent or degree.
• the term “substantially” means that the subsequently described event or circumstance occurs at least 80% of the time, or at least 85% of the time, or at least 90% of the time, or at least 95% of the time.
• the term “substantially adjacent” may mean that two items are 100% adjacent to one another, or that the two items are within close proximity to one another but not 100% adjacent to one another, or that a portion of one of the two items is not 100% adjacent to the other item but is within close proximity to the other item.
• threshold value refers to a value (or values) that are used as a reference to attain information on and/or classify the results of a measurement, for example, the results of a measurement attained in an assay.
• a threshold value can be determined based on one or more control samples.
• a threshold value can be determined prior to, concurrently with, or after the measurement of interest is taken.
• a threshold value can be a range of values.
• a threshold value can be a value (or range of values) reported in the relevant field (e.g., a value found in a standard table).
• treatment includes the application or administration of a compound to a subject with the purpose of delaying, slowing, stabilizing, curing, healing, alleviating, relieving, altering, remedying, less worsening, ameliorating, improving, or affecting the disease or condition, the symptom of the disease or condition, or the risk of (or susceptibility to) the disease or condition.
• treating refers to any indication of success in the treatment or amelioration of an injury, pathology or condition, including any objective or subjective parameter such as abatement; remission; lessening of the rate of worsening; lessening severity of the disease; stabilization, diminishing of symptoms or making the injury, pathology or condition more tolerable to the subject; slowing in the rate of degeneration or decline; making the final point of degeneration less debilitating; or improving a subject's physical or mental well-being.
• the term “treating” means reducing or ameliorating the progression, severity, and/or duration of COVID-19, or ameliorating one or more symptoms of COVID-19 caused by administration of one or more therapies (e.g., one or more therapeutic agents).
• the term “treatment” means to ameliorate measurable physical parameters of COVID-19.
• the term “treating” means reducing or ameliorating the progression, severity, and/or duration of a viral, bacterial, or fungal infection or ameliorating one or more symptoms of a viral, bacterial, or fungal infection caused by administration of one or more therapies (e.g., one or more therapeutic agents).
• the term “treatment” means to ameliorate measurable physical parameters of a viral, bacterial, or fungal infection.
• the term “treating” means reducing or ameliorating the progression, severity, and/or duration of sepsis, or ameliorating one or more symptoms of a sepsis caused by administration of one or more therapies (e.g., one or more therapeutic agents).
• the term “treatment” means to ameliorate measurable physical parameters of sepsis.
• “treating” changes the natural or presenting state of a subject.
• the term “therapeutically effective amount” means the amount of compound that, when administered to a subject for treating or preventing a particular disorder, disease or condition, is sufficient to effect such treatment or prevention of that disorder, disease or condition.
• Dosages and therapeutically effective amounts may vary for example, depending upon a variety of factors including the activity of the specific agent employed, the age, body weight, general health, gender, and diet of the subject, the time of administration, the route of administration, the rate of excretion, and any drug combination, if applicable, the effect which the practitioner desires the compound to have upon the subject and the properties of the compounds (e.g., bioavailability, stability, potency, toxicity, etc.), and the particular disorder(s) the subject is suffering from.
• the therapeutically effective amount that is administered intravenously may depend on the subject's blood parameters e.g., lipid profile, insulin levels, glycemia or liver metabolism.
• the therapeutically effective amount will also vary according to the severity of the disease state, organ function, or underlying disease or complications. Such appropriate doses may be determined using any available assays. When one or more of the compounds or therapeutic agents is to be administered to humans, a physician may for example, prescribe a relatively low dose at first, subsequently increasing the dose until an appropriate response is obtained.
• the term “viral disease” refers to the pathological state resulting from the presence of a virus in a cell or subject or the invasion of a cell or subject by a virus.
• influenza virus disease refers to the pathological state resulting from the presence of an influenza (e.g., influenza A or B virus) virus in a cell or subject or the invasion of a cell or subject by an influenza virus. In specific embodiments, the term refers to a respiratory illness caused by an influenza virus.
• COVID-19 disease refers to the pathological state resulting from the presence of SARS-CoV-2 (e.g., a SARS-CoV-2 variant) virus in a cell or subject or the invasion of a cell or subject by a SARS-CoV-2 virus or variant thereof.
• the term refers to a respiratory illness caused by a corona virus such as SARS-CoV-2 or variant thereof.
• a corona virus such as SARS-CoV-2 or variant thereof.
• the term “sepsis” has been used to describe a variety of clinical conditions related to systemic manifestations of inflammation accompanied by an infection. Because of clinical similarities to inflammatory responses secondary to non-infectious etiologies, identifying sepsis has been a particularly challenging diagnostic problem.
• Sepsis definitions can change over time and can also vary across various clinical/hospital systems (e.g., Singer M, Deutschman CS, Seymour CW, Shankar- Hari M, Annane D, Bauer M, Bellomo R, Bernard GR, Chiche JD, Coopersmith CM, Hotchkiss RS.
• Bone RC, Balk RA, Cerra FB et al. American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference: definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis.
• FIG.1 is a block diagram of system 100 in which an embodiment can be implemented, for example, as a system 100 for predicting presence for a disease, configured to perform the process as described therein.
• said system 100 includes a processing unit 101, a memory 102, a storage unit 103, an input unit 104, a bus 106, an output unit 105, and a network interface 107.
• the processing unit 101 means any type of computational circuit, such as, but not limited to, a microprocessor, microcontroller, complex instruction set computing microprocessor, reduced instruction set computing microprocessor, very long instruction word microprocessor, explicitly parallel instruction computing microprocessor, graphics processor, digital signal processor, or any other type of processing circuit.
• the processing unit 101 may also include embedded controllers, such as generic or programmable logic devices or arrays, application specific integrated circuits, single-chip computers, and the like.
• the memory 102 may be volatile memory and non-volatile memory. The memory 102 may be coupled for communication with said processing unit 101.
• the processing unit 101 may execute instructions and/or code stored in the memory 102.
• the memory 102 may include any suitable elements for storing data and machine-readable instructions, such as read only memory, random access memory, erasable programmable read only memory, electrically erasable programmable read only memory, a hard drive, a removable media drive for handling compact disks, digital video disks, diskettes, magnetic tape cartridges, memory cards, and the like.
• the memory 102 includes a disease prediction module 110 stored in the form of machine-readable instructions on any of said above-mentioned storage media and may be in communication to and executed by processor 101. When executed by the processor 101, the disease prediction module 110 causes the processor 101 to predict presence of a disease in a subject.
• the storage unit 103 may be a non-transitory storage medium which stores a medical database 112.
• the medical database 112 is a repository of patient data, including blood parameters, that is maintained by a healthcare service provider.
• the input unit 104 may include input means such as keypad, touch-sensitive display, camera (such as a camera receiving gesture-based inputs), etc. capable of receiving in-put signal.
• the bus 106 acts as interconnect between the processing unit 101, the memory 102, the storage unit 103, the input unit 104, the output unit 105 and the network interface 107.
• the system 100 in accordance with an embodiment of the present disclosure includes an operating system employing a graphical user interface. Said operating system permits multiple display windows to be presented in the graphical user interface simultaneously with each display window providing an interface to a different application or to a different instance of the same application.
• a cursor in said graphical user interface may be manipulated by a user through a pointing device.
• the position of the cursor may be changed and/or an event such as clicking a mouse button, generated to actuate a desired response.
• One of various commercial operating systems such as a version of Microsoft WindowsTM, a product of Microsoft Corporation located in Redmond, Washington may be employed if suitably modified. Said operating system is modified or created in accordance with the present disclosure as described.
• non-transitory computer readable media containing executable instructions that when executed cause a processor to perform operations including a method as provided herein are provided.
• the present disclosure includes an article of manufacture, such as a system or component thereof including a non-transitory computer-readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method including: predicting a presence of a disease in accordance with the present disclosure.
• the present disclosure includes non- transient computer readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method including: predicting a presence of a disease in accordance with the present disclosure.
• FIG.2 illustrates a process flow 200 for predicting presence of a disease in a subject, according to another embodiment.
• a plurality of parameters associated with the subject is received.
• the parameters associated with the subject may be obtained from a complete blood count/differential test performed for the subject.
• the parameters may include red blood cells (RBC) parameters, white blood cells (WBC) and platelet parameters.
• the RBC parameters include hemoglobin level, hematocrit, RBC size, hemoglobin level in individual RBC (normal, high low), mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs with hemoglobin concentrations ⁇ 28g/dL and ⁇ 41g/dL, number of RBCs with hemoglobin volumes ⁇ 60 fL and ⁇ 120 fL hemoglobin distribution width, RBC mean corpuscular volume, RBC volume distribution width, total RBC count, hemoglobin content distribution width and mean of hemoglobin content.
• the RBC size- based parameters include RBC counts below 60 fL and above 120 fL. Additional parameters such as cell hemoglobin concentration mean, number of RBCs with hemoglobin concentrations greater than 41 g/dL, and number of RBCs with hemoglobin concentrations less than 28 g/dL are also considered.
• the parameters associated with the subject may be obtained from a complete blood count/differential test performed for the subject.
• the parameters may include red blood cells (RBC) parameters, white blood cells (WBC) and platelet parameters.
• the RBC parameters are preselected and include one or more of hemoglobin level, hematocrit, RBC size, hemoglobin level in individual RBC (normal, high low), mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs with hemoglobin concentrations ⁇ 28g/dL and ⁇ 41g/dL, number of RBCs with hemoglobin volumes ⁇ 60 fL and ⁇ 120 fL hemoglobin distribution width, RBC mean corpuscular volume, RBC volume distribution width, total RBC count, hemoglobin content distribution width and mean of hemoglobin content, and combinations thereof.
• the WBC parameters include WBC type such as lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, Large Unstained Cells (LUC), blasts and platelets. Additionally, WBC count, WBC percentage, ratio of neutrophils to lymphocytes, ratio of large unstained cells to lymphocytes, number of WBCs indicating pseudobasophilia, and WBC maturity parameters such as Delta Neutrophil Index are also considered.
• the parameters associated with the subject are indicative of inflammation and immune activation. Hence the parameters may be used to determine the presence of an infection in the subject. In embodiments, the WBC parameters are preselected.
• the WBC parameters include one or more of WBC type (such as lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, Large Unstained Cells (LUC), blasts and platelets, WBC count, WBC percentage, ratio of neutrophils to lymphocytes, ratio of large unstained cells to lymphocytes, number of WBCs indicating pseudobasophilia, WBC maturity parameters such as Delta Neutrophil Index, and combinations thereof.
• WBC type such as lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, Large Unstained Cells (LUC), blasts and platelets
• WBC count WBC percentage
• ratio of neutrophils to lymphocytes ratio of large unstained cells to lymphocytes
• number of WBCs indicating pseudobasophilia such as Delta Neutrophil Index, and combinations thereof.
• platelet parameters are included and considered. Non- limiting platelet parameters include those shown in a plate
• a threshold or threshold value for sensitivity and specificity associated with a trained machine learning model is determined for the desired use case.
• the threshold value is predetermined.
• the trained machine learning model is an ensemble of decision tree-based methods. For example, Extra Trees Classifier and Light Gradient Boosting Machine (LightGBM) may be used as the machine learning models.
• the ensemble of decision tree-based models is trained using hematology data associated with a plurality of subjects.
• the hematology data includes the WBC and RBC data obtained from Complete Blood Count/Differential test associated with the plurality of subjects.
• the machine learning models are trained with 5-fold cross validation and optimized for sensitivity.
• Extra Trees Classifier is an ensemble learning technique that improves accuracy and controls over- fitting by learning randomized decision trees on different data subsamples.
• Light Gradient Boosting Machine (LightGBM) is a boosting approach for decision tree learning in which the continuous feature values are discretized into bins for faster training and reduced memory usage.
• the different classifiers with the associated hyperparameters used in the ensemble are specified below (a-g).
• the weights for each classifier in the ensemble in the same order as listed [2, 1, 1, 2, 1, 1, 1]. a.
• the requirement scenario associated with the healthcare provider includes at least one of reduction of RT-PCR testing burden, reduction of infectious disease testing burden, substitution of RT-PCR testing, and/or determination of a need for testing of subjects.
• the algorithm can be used at a reasonably high sensitivity (0.80-0.95), such that most positives will be predicted as positives and some negatives will be predicted as positives.
• RT-PCR testing can be avoided in samples that are predicted negative by the algorithm. This approach can significantly reduce the RT-PCR testing burden in many labs while missing a small number of positives. This use case can be useful in the pandemic scenario.
• the algorithm can be used at a very high sensitivity (0.99-1.0), such that all positives will be predicted positive, and many negatives will be predicted as positives.
• RT-PCR testing can be avoided in samples that are predicted negative, without missing any true positives. This approach can reduce the RT-PCR testing burden in many labs/hospitals without missing any positives. This use case can be useful in the pandemic scenario.
• the algorithm can be used at a very high specificity (0.99-1.0) such that all negatives will be predicted as negatives, but some positives will be flagged as positives.
• the presence of the disease in the subject is predicted using the trained machine learning model. Such prediction is made based on the sensitivity and specificity of the disease and the plurality of parameters associated with the subject. Binary classification models generate the probability of the input to be classified into categories (in the present case, COVID positive or negative). The final positive/negative callout is based on the threshold chosen.
• the default threshold in such models may be 0.5 (i.e., if the model outputs a probability of positive prediction as ⁇ 0.5 for any input, the input is labeled positive, and if probability of positive prediction is ⁇ 0.5, then it is labeled negative). Adjusting the threshold to a different value depending on the scenario enables curating the output from the model. For example, if a higher positive prediction rate (fewer false positives) is required, a higher threshold for sensitivity and specificity is chosen. However, if a higher negative prediction is required (fewer false negatives), a lower threshold for sensitivity and specificity is chosen.
• the prediction is outputted on an output unit 105. [0059] For example, the prediction may be displayed on a display unit.
• the method includes determining a need for hospitalization of the subject based on the plurality of the parameters associated with the subject, using the trained machine learning model.
• the model is trained using the parameters associated with the plurality of subjects tested positive for the disease, wherein the parameters associated with the plurality of subjects is obtained within 2-3 days of the subjects testing positive for the disease. Additionally, the model is also trained with health data associated with the subjects that may be observed and collected over 7-10 days after the subjects testing positive for the disease.
• the method may include predicting an occurrence of long COVID-19 in the subject. The model is trained using the parameters associated with the plurality of subjects tested positive for the disease and health data associated with the subjects that may be observed and collected over the next months after the subjects testing positive for the disease.
• FIG.3 illustrates a receiver operating characteristic curve, or ROC curve 300 for choosing threshold for sensitivity and specificity associated with a trained machine learning model for a requirement scenario, according to an embodiment.
• the ROC curve is a graphical plot that illustrates the diagnostic ability of a binary classification model as its discrimination threshold is varied. For various thresholds, ROC plots the true positive rate (sensitivity) against the false positive rate (probability of false alarm/ 1-specificity). The diagonal divides the ROC space. Points above the diagonal represent good classification results (better than random); points below the line represent bad results (worse than random).
• the present disclosure includes a method of treating a subject in need thereof by determining that the subject is a subject in need thereof (such as has a pathogenic infection, or sepsis), and subsequently treating the subject.
• the methods of the present disclosure can be applied to diagnose a patient as positive for a pathogenic infection, sepsis, a viral infection such as COVID- 19, followed by administering a therapeutic agent or compound to the subject in need thereof in a therapeutically effective amount.
• a physician may administer a therapeutically effective amount of any drug, therapeutic agent or biologic suitable for treating the pathogenic infection, or disease state subsequent to diagnosis in accordance with the present disclosure.
• agent suitable for treating a viral infection such as COVID-19 includes VEKLURY® brand remdesivir (100 MG for injection antiviral medication for treatment of patients hospitalized with COVID-19). See also, e.g., the international patent application WO 2017/049060 A1, and US US20210395345 entitled Methods for treating or preventing sars-cov-2 infections and covid-19 with anti-sars-cov-2 spike glycoprotein antibodies.
• the therapeutically effective amount of agent provided may be determined by a physician based on the subject’s response, comorbidities, and the like.
• a therapeutic compound is administered in an effective amount to a subject in a diseased state.
• a therapeutically effective amount of a compound is administered to a subject in need thereof in an amount sufficient to change the natural or presenting state of the subject.
• a subject may be in a diseased state due to infection with an RNA virus.
• Non-limiting examples of an RNA virus include segmented, single- stranded, negative sense RNA viruses (e.g., Orthomyxoviruses); non-segmented, single-stranded, negative sense RNA viruses (Mononegavirales); non-segmented, single-stranded, positive sense RNA viruses (e.g., Coronaviruses); ambisense RNA viruses (e.g., Bunyavirus and Arenavirus, e.g., Lassa virus, Junin virus, Machupo virus, and lymphocytic choriomeninfitis virus); and double-stranded RNA viruses (e.g., Reoviruses).
• negative sense RNA viruses e.g., Orthomyxoviruses
• non-segmented, single-stranded, negative sense RNA viruses Mononegavirales
• non-segmented, single-stranded, positive sense RNA viruses e.g., Coronaviruses
• viruses causing disease include a segmented, single-stranded, negative sense RNA viruses; a non-segmented, single-stranded, negative sense RNA virus; a non-segmented, single-stranded, positive sense RNA viruses; or a double- stranded RNA viruses (e.g., Reoviruses).
• Additional RNA viruses include: Picornaviruses, togaviruses (e.g., Sindbis virus), flaviviruses, and coronaviruses. The virus can be any type, species, and/or strain of picornavirus, togavirus flavivirus, and coronavirus.
• Further disease-causing viruses includes: respiratory syncytial virus (RSV), influenza virus (influenza A virus, influenza B virus, or influenza C virus), human metapneumovirus (HMPV), rhinovirus, parainfluenza virus, SARS Coronavirus, human immunodeficiency virus (HIV), hepatitis virus (A, B, C), ebola virus, herpes virus, rubella, variola major, and variola minor.
• RSV respiratory syncytial virus
• influenza virus influenza A virus, influenza B virus, or influenza C virus
• HMPV human metapneumovirus
• rhinovirus parainfluenza virus
• SARS Coronavirus SARS Coronavirus
• HAV human immunodeficiency virus
• HAV human immunodeficiency virus
• hepatitis virus A, B, C
• ebola virus herpes virus
• rubella variola major
• variola minor variola minor.
• a therapeutic composition is administered to a subject who has been diagnosed with a disease caused by infection with a virus, e.g., the patient has been infected by respiratory syncytial virus (RSV), influenza virus (influenza A virus, influenza B virus, or influenza C virus), human metapneumovirus (HMPV), rhinovirus, parainfluenza virus, SARS Coronavirus, human immunodeficiency virus (HIV), hepatitis virus (A, B, C), ebola virus, herpes virus, rubella, variola major, and/or variola minor.
• RSV respiratory syncytial virus
• influenza virus influenza A virus, influenza B virus, or influenza C virus
• HMPV human metapneumovirus
• rhinovirus parainfluenza virus
• SARS Coronavirus human immunodeficiency virus
• HAV human immunodeficiency virus
• hepatitis virus A, B, C
• ebola virus herpes virus
• rubella rubella
• Non-limiting examples of disease-causing bacteria include: Streptococcus pneumoniae, Mycobacterium tuberculosis, Chlamydia pneumoniae, Bordetella pertussis, Mycoplasma pneumoniae, Haemophilus influenzae, Moraxella catarrhalis, Legionella, Pneumocystis jiroveci, Chlamydia psittaci, Chlamydia trachomatis, Bacillus anthracis, and Francisella tularensis, Borrelia burgdorferi, Salmonella, Yersinia pestis, Shigella, E. coli, Corynebacterium diphtheriae, and Treponema pallidum.
• a composition is administered to a patient who has been diagnosed with a disease caused by infection with a bacteria, e.g., the patient has been infected by Streptococcus pneumoniae, Mycobacterium tuberculosis, Chlamydia pneumoniae, Bordetella pertussis, Mycoplasma pneumoniae, Haemophilus influenzae, Moraxella catarrhalis, Legionella, Pneumocystis jiroveci, Chlamydia psittaci, Chlamydia trachomatis, Bacillus anthracis, and Francisella tularensis, Borrelia burgdorferi, Salmonella, Yersinia pestis, Shigella, E.
• a composition is administered to a patient diagnosed with a disease caused by a fungal infection, e.g., the patient has been infected by Blastomyces, Paracoccidioides, Sporothrix, Cryptococcus, Candida, Aspergillus, Histoplasma, Cryptococcus, Bipolaris, Cladophialophora, Cladosporium, Drechslera, Exophiala, Fonsecaea, Phialophora, Xylohypha, Ochroconis, Rhinocladiella, Scolecobasidium, and/or Wangiella.
• a composition is administered to a patient who has been diagnosed with a disease caused by infection with a yeast, e.g., the patient has been infected by Aciculoconidium, Botryoascus, Brettanomyces, Bullera, Bulleromyces, Candida, Citeromyces, Clavispora, Cryptococcus, Cystofilobasidium, Debaromyces, Debaryomyces, Dekkera, Dipodascus, Endomyces, Endomycopsis, Erythrobasidium, Fellomyces, Filobasidium, Guilliermondella, Hanseniaspora, Hansenula, Hasegawaea, Hyphopichia, Issatchenkia, Kloeckera, Kluyveromyces, Komagataella, Leucosporidium, Lipomyces, Lodderomyces, Malassezia- Mastigomyces, Metschnikowia, Mrakia,
• a composition is administered to a subject infected with a parasite, e.g., the patient has been infected by Babesia, Cryptosporidium, Entamoeba histolytica, Leishmania, Giardia lamblia, Plasmodium, Toxoplasma, Trichomonas, Trypanosoma, Ascaris, Cestoda, Ancylostoma, Brugia, Fasciola, Trichinella, Schistosoma, Taenia, Cimicidae, Pediculus, and/or Sarcoptes. See additional information in U.S. patent publication no.20170321192 (herein entirely incorporated by reference).
• Embodiment 1 A system (100) for predicting presence of a disease in a subject, the system (100) including: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) including patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105).
• a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105).
• Embodiment 2 A system (100) for predicting presence of a disease in a subject, the system (100) including: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the threshold or threshold value for sensitivity and/or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider.
• Embodiment 3 A system (100) for predicting presence of a disease in a subject, the system (100) including: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold or threshold value of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the threshold or threshold value for sensitivity and/or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider, and wherein the requirement scenario associated with the healthcare provider includes at least one of reduction of RT-PCR testing burden, substitution of RT-PCR testing, and/or determination of a need for testing of subjects.
• Embodiment 4 A system (100) for predicting presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the plurality of parameters associated with the subject comprises red blood cells (RBC) parameters and white blood cells (WBC) parameters.
• RBC red blood cells
• WBC white blood cells
• Embodiment 5 A system (100) for predicting presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the plurality of parameters associated with the subject comprises red blood cells (RBC) parameters and white blood cells (WBC) parameters, and wherein the RBC parameters comprise one or more of hemoglobin level, hematocrit, RBC size, hemoglobin level in individual RBC, mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs with
• Embodiment 6 A system (100) for predicting presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the plurality of parameters associated with the subject comprises red blood cells (RBC) parameters and white blood cells (WBC) parameters, wherein the WBC parameters comprise one or more of WBC type, WBC count, WBC percentage, ratio of neutrophils to lymphocytes, ratio of large unstained cells to lymphocytes, number of WBCs indicating pseudobasophilia, and WBC maturity parameters
• Embodiment 7 A system (100) for predicting presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the plurality of parameters associated with the subject comprises red blood cells (RBC) parameters and white blood cells (WBC) parameters, and wherein the RBC parameters comprise one or more of hemoglobin level, hematocrit, RBC size, hemoglobin level in individual RBC, mean corpuscular volume, mean corpuscular hemoglobin concentration, normal RBCs, number of RBCs with
• Embodiment 8 A system (100) for predicting presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the trained machine learning model is an ensemble of decision tree- based methods.
• a system (100) for predicting presence of a disease in a subject comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the disease prediction module (110) is further configured to predict a need for hospitalization of the subject based on the plurality of the parameters associated with the subject, using the trained machine learning model.
• Embodiment 10 A system (100) for predicting presence of a disease in a subject, the system (100) comprising: one or more processing units (101); a medical database (112) coupled to the one or more processing units (101), the medical database (112) comprising patient data; and a disease prediction module (110) configured to: receive a plurality of parameters associated with the subject; determine a threshold for sensitivity and/or specificity associated with a trained machine learning model; predict using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and output the prediction on an output unit (105), wherein the disease prediction module (110) is further configured to predict occurrence of long COVID-19 in the subject.
• Embodiment 11 A method (200) of predicting presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit.
• a method (200) of predicting presence of a disease in a subject comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, wherein the threshold for sensitivity and/or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider.
• a method (200) of predicting presence of a disease in a subject comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, wherein the threshold for sensitivity and/or specificity associated with the trained machine learning model is determined based on a requirement scenario associated with a healthcare provider, and wherein the requirement scenario associated with the healthcare provider comprises at least one of reduction of RT-PCR testing burden, substitution of RT-PCR testing, and/or determination of a need for testing of subjects.
• Embodiment 14 A method (200) of predicting presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, wherein the disease is COVID-19.
• Embodiment 15 Embodiment 15.
• a method (200) of predicting presence of a disease in a subject comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, further comprising: predicting a need for hospitalization of the subject based on the plurality of the parameters associated with the subject, using the trained machine learning model; and predicting an occurrence of long COVID-19 in the subject. [0085] Embodiment 16.
• a method (200) of predicting presence of a disease in a subject comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, wherein the disease is COVID-19.
• a method (200) of predicting presence of a disease in a subject comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, further comprising: predicting a need for hospitalization of the subject based on the plurality of the parameters associated with the subject, using the trained machine learning model; and predicting an occurrence of viral, bacterial, or fungal infection in the subject. [0087] Embodiment 18.
• An article of manufacture such as a system or component thereof including a non-transitory computer-readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method of Embodiments 15-17 above. [0088] Embodiment 19.
• the present disclosure includes non-transient computer readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method comprising: predicting presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, further comprising: predicting a need for hospitalization of the subject based on the plurality of the parameters associated with the subject, using the trained machine learning model; and predicting an occurrence of viral, bacterial, or fungal infection in the subject.
• Embodiment 20 includes non-transient computer readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method comprising: predicting presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, further comprising: predicting a need for hospitalization of the subject based on the plurality of the parameters associated with the subject, using the trained machine learning model; and predicting an occurrence of sepsis in the subject.
• Embodiment 21 An article of manufacture including a non- transitory computer-readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method comprising: predicting a presence of a disease in accordance with an embodiment of the present disclosure.
• Embodiment 22 A non-transient computer readable medium with instructions encoded thereon, the instructions configured to cause one or more processors to perform a method comprising: predicting a presence of a disease in accordance with at least one embodiment of the present disclosure.
• Embodiment 23 A method of treating a subject in need thereof including predicting a presence of a disease in a subject in accordance with an embodiment of the present disclosure and subsequently treating the subject in need thereof.
• treating includes administering to the subject a therapeutically effective amount of a composition.
• Embodiment 24 A method of treating a subject in need thereof including predicting presence of a disease in a subject, the method (200) comprising: receiving a plurality of parameters associated with the subject; determining a threshold for sensitivity and/or specificity associated with a trained machine learning model; predicting using the trained machine learning model the presence of the disease in the subject based on a specific threshold of the model and the plurality of parameters associated with the subject; and outputting the prediction on an output unit, and subsequently treating the subject in need thereof.
• treating includes administering to the subject a therapeutically effective amount of a composition.
• the disease is characterized as viral disease, bacterial disease, COVID-19, long-Covid, or the like.
• the present disclosure includes a system for predicting a presence of a disease in a population.
• the system includes a processing unit and a memory.
• the memory includes a disease prediction module configured to receive a plurality of parameters associated with one or more subjects and determine a threshold for sensitivity and/or specificity associated with a trained machine learning model. Additionally, the module is configured to predict using the trained machine learning model the presence of the disease in one or more subjects or a population of subjects and the plurality of parameters associated with the subject or population of subjects and output the prediction on an output unit.
• present example describes a COVID-19 detection algorithm based on CBC/Diff hematology data.
• present disclosure includes a multiparametric machine learning algorithm to run on, among other things, a hematology instrument and provide a COVID-19 flag along with the CBC/Diff data.
• the algorithm can be run at different sensitivity and specificity values based on the intended use case, in a pandemic or an endemic situation.
• the training data used in the algorithm is based on an unvaccinated cohort from 2020. Additional data from more varied positive and negative individuals may be used to improve the robustness of the algorithm.
• CBC/Diff Complete Blood Count with Differential
• the present disclosure combines CBC/Diff results from the Siemens Healthineers ADVIA® 2120i Hematology System with a multiparametric machine learning algorithm for rapid flagging of COVID-19 (Coronavirus Disease 2019) positivity. Diagnostic data generation and running the algorithm on the data all happen on the same instrument, ensuring easy data access. Additionally, the algorithm can also run on a connected lab information system. Such an algorithm can be valuable when access to COVID testing is limited and for predicting COVID-19 in unsuspected individuals thus prompting a confirmatory test. Here, various pandemic and endemic applications for this algorithm are demonstrated. Methods.
• CBC/Diff data for 309 COVID-19 positives and 245 negatives were obtained from a 2020 pre-vaccinated cohort from Mumbai, India. Using this data, machine learning classification models were trained with 5-fold cross-validation for distinguishing COVID-19 positives from negatives. Results.
• the final ML model was an ensemble of decision tree-based methods, with an accuracy of 74% and AUROC of 0.79 on the test set. The most significant features include eosinophil and lymphocyte counts that have been reported earlier. The model was able to predict hospitalized COVID-19 patients more accurately (92% correctly classified) than non-hospitalized patients (70% correctly classified). Conclusion. A predictive algorithm, method, and system is provided for rapid COVID- 19 flagging using data from routine CBC testing.
• Hematology data provides a snapshot of the state of health of a patient and represents the host’s response to an infection. Additionally, a hematology test such as a CBC/Diff is commonly prescribed before any disease suspicion and is thus an ideal platform for flagging and stratifying a possible infection or a health condition, thus enabling an earlier intervention and streamlining different medical resources (See e.g., Barnes PW, McFadden SL, Machin SJ, Simson E. The international consensus group for hematology review: suggested criteria for action following automated CBC and WBC differential analysis.
• Multiparametric machine learning (ML) algorithms enable modeling of complex relationships between multiple predictive parameters and generate predictions more accurately and efficiently.
• ML machine learning
• the present disclosure uses multiple hematology parameters measured/derived using a single instrument to make predictions.
• an algorithm with high negative predictive value could reduce the number of patients undergoing testing before procedures like surgery, thereby reducing time to patient care.
• an algorithm with high specificity could be used for flagging asymptomatic individuals who can then be tested to confirm the infection.
• the same ML algorithm with different classification thresholds can be used in multiple scenarios described as above.
• a hematology based multiparametric ML algorithm could be used for COVID flagging within seconds to minutes. The reduced time to result can avoid unnecessary quarantining of negatives and rapid access to care for patients that need to undergo procedures like surgery in a pandemic scenario and also be used for flagging unsuspected positive during an endemic scenario.
• the present disclosure includes a multi-parametric machine learning algorithm to predict COVID-19 positivity in individuals using CBC/Diff data derived from a single instrument.
• the multiparametric ML algorithm predicts COVID-19 positivity in individuals with an accuracy of 74% and AUROC (area under the receiver operating characteristic curve) of 0.79.
• the algorithm predicts positive hospitalized patients with 91.7% accuracy, giving us confidence in our algorithm performance and the potential of the parameters selected.
• Materials and methods Datasets [0102] CBC/Diff data was obtained from two ADVIA® 2120i instruments installed in Dr. Jariwala Laboratory & Diagnostics in Mumbai (India) for this analysis.
• the present disclosure includes a retrospective study on a cohort of 309 positive and 245 COVID- 19 negative individuals, determined using RT-PCR tests. Only anonymized data was sent to the researchers, that included information on the sex and age of the individuals and the disease outcome in COVID-19 positive individuals (whether they were hospitalized / admitted to the ICU / survived or died).
• the COVID-19 positives and negatives were concurrently measured during the first wave of infection in India (May 2020–Jan 2021) before any vaccination drives.
• the COVID-19 negatives included in this study were individuals receiving medical attention for other conditions during the first wave of the pandemic.
• 60 were hospitalized at the time of CBC/Diff testing, out of which 29 were in the ICU and 53 survived the infection.
• the demographics and blood cell counts of COVID-19 positive and negative individuals included in this study are summarized in Table 1. TABLE 1 Demographics and blood cell counts of COVID-19–positive and negative individuals used in this study.
• Parameter Selection [0103] The starting dataset included all the parameters included in the ADVIA 2120i CBC/Diff raw data set. Parameters were eliminated using various criteria, e.g., where the median value did not vary significantly between the positives and the negatives (done using box plots and p-values for Kruskal-Wallis H-test), parameters that measured only instrument characteristics, etc. Redundant and highly correlated parameters were removed (Spearman correlation coefficient > 0.9).
• an excellent model for the dataset is a voting ensemble of multiple tree- based classifiers.
• the different classifiers with the associated hyperparameters used in the ensemble are specified below (a-g).
• the weights for each classifier in the ensemble in the same order as listed are [2, 1, 1, 2, 1, 1, 1].
• MIMIC LightGBM explainer approximates the complex, black-box ensemble model obtained from Azure AutoML using a LightGBM surrogate model to obtain feature/parameter importance values.
• a comparison of the results from our 48- parameter model with other models learned using the most important parameters was performed. Different use cases for the algorithm.
• Model evaluation [0107] The entire dataset was divided into separate training (80% of total data) and test sets (remaining 20% data). The model was trained using 5-fold cross- validation on the training set and the model performance was evaluated on the test set. The test data set consisted of 62 positives and 49 negatives.
• FIG.4 shows the ROC (Receiver operating curve) curve computed using the test data. Since the positives and negatives in the test are balanced, AUROC is a good performance metric for validation. [0108] Out of 62 positives in the test set, 12 were hospitalized whereas the remaining 50 were not. It was observed that the hospitalized positives were more accurately predicted than non-hospitalized positives (Table 2). TABLE 2 [0109] This gives increased confidence in the selected parameters.
• the algorithm can be used at a sensitivity of 0.94 and specificity of 0.39 (point 1 in the ROC curve) to make predictions for this use case. Depending on the sensitivity chosen, the algorithm can capture a large majority of the positives. This result could be used as an additional point in clinical decision making in the absence of a confirmatory test.
• the algorithm could be used at high sensitivity to reduce infection testing burden. This could be especially useful in high volume labs.
• the rapid turn-around time of CBC/Diff results on the ADVIA 2120i analyzer (about a minute) along with algorithm running time (a few seconds) provides a faster alternative to RT-PCR or antigen testing.
• the algorithm has an NPV of 1, where negatives are predicted with high certainty and need not be tested and can be exempted from quarantine. Individuals predicted positive by the algorithm will need to be followed up by COVID specific testing. This approach can reduce testing burden by about 10%.
• the algorithm could also be used in an endemic scenario to drive testing in individuals with fever in the emergency department (ED), when COVID-19 is not suspected.
• Embodiments provide a multi-parametric machine learning algorithm for flagging COVID-19 positivity in individuals using parameters obtained from a CBC/Diff test performed on the Siemens Healthineers ADVIA® 2120i Hematology System.
• a CBC/Diff test is a routinely done blood test and can help detect various health conditions.
• Embodiments include an ensemble of tree-based algorithms to predict COVID-19 positive or negative status using hematology data.
• a classifier predicts COVID-19 positivity with an accuracy of 74% on the entire test set and an accuracy of 91.7% on hospitalized patients. Better performance on hospitalized individuals gives excellent confidence in the parameter selection process of the present disclosure.
• an algorithm embodiment can advantageously be used in different settings in pandemic and endemic scenarios, including its use as a faster alternative to RT-PCR testing, reducing testing burden in a pandemic and detecting COVID-19 in unsuspected, asymptomatic individuals.
• Another study See Kukar M, Gun ⁇ ar G, Vovko T, Podnar S, ⁇ ernel ⁇ P, Brvar M, Zalaznik M, et al. COVID-19 diagnosis by routine blood tests using machine learning. Scientific reports 2021;11:1-9) has used routine blood parameters for COVID-19 diagnosis using a machine learning algorithm, but those parameters need to be collected from multiple different tests, in contrast to embodiments of the present disclosure using parameters obtained from a single instrument.
• Kukar et al includes a skewed dataset (3% positives, 97% negatives). Though the prevalence rate of COVID-19 at the time of that publication was 3%, it is believed that AUROC is not the right metric for a heavily skewed data set, and perhaps some other metric like the area under the precision-recall curve (AUPRC) could have been a better measure to report cross-validation results. On the other hand, the present disclosure includes a hold-out test dataset (20% of total data) with balanced number of positives and negatives to report AUROC and other metrics.
• the algorithm used described herein includes a total of 48 parameters including percentages of various white cell types, like lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, large unstained cells (LUC) and blasts.
• various white cell types like lymphocytes, neutrophils, basophils, monocytes, eosinophils, polymorphonuclear cells, large unstained cells (LUC) and blasts.
• the parameter set also included platelet count and several red blood cell (RBC) parameters like hemoglobin level, mean corpuscular volume (MCV), cell hemoglobin concentration mean (g/dL), the number of red cells with low, normal and high hemoglobin concentrations and volumes, hemoglobin distribution width (HDW) which is the standard deviation of the hemoglobin concentration (HC) histogram and is expressed in g/dL, red blood cell distribution width (RDW), total RBC counts.
• RBC red blood cell
• the parameter set also included some raw instrument measurements that are unique to the ADVIA 2120i analyzer.
• neutrophil-to-lymphocyte ratio NLR
• LUC-to-lymphocyte ratio LUC-to-lymphocyte ratio
• CCLM Clinical chemistry and laboratory medicine 2020; 58:1131-4; and Zhang J-j, Dong X, Cao Y-y, Yuan Y-d, Yang Y-b, Yan Y-q, Akdis CA, et al. Clinical characteristics of 140 patients infected with SARS-CoV-2 in Wuhan, China. Allergy 2020;75:1730-41), lymphocytes have been observed to be an important biomarker in diagnosis and as well as prognosis of COVID-19 (See e.g., Henry BM, De Oliveira MHS, Benoit S, Plebani M, Lippi G.
• Leukopenia was observed in 33.7% of hospitalized patients on admission (See e.g., Guan W-j, Ni Z-y, Hu Y, Liang W-h, Ou C-q, He J-x, Liu L, et al. Clinical characteristics of coronavirus disease 2019 in China. New England journal of medicine 2020;382:1708- 20).
• the prevalence of thrombocytopenia has been widely reported in COVID-19 patients, particularly in more severe patients (See e.g., Henry BM, De Oliveira MHS, Benoit S, Plebani M, Lippi G. Hematologic, biochemical and immune biomarker abnormalities associated with severe illness and mortality in coronavirus disease 2019 (COVID-19): a meta-analysis ⁇ .
• test set consists of 43 males and 19 females in the COVID-19 positive cohort, and 22 males and 27 females in the negative cohort, respectively.
• model performance is consistent across males and females, with around 74% prediction accuracy for both thus suggesting no sex related bias in prediction (Supplemental Table 1).
• prediction accuracy of embodiments of the present disclosure was compared across the test set for different age groups (Supplemental Table 2).
• the present disclosure combines CBC/Diff results from the Siemens Healthineers ADVIA® 2120i Hematology System with a multiparametric machine learning algorithm for rapid flagging of pathogenic disease positivity such as viral, bacterial, or fungal disease positivity. Diagnostic data generation and running the algorithm on the data occur on the same instrument, to ensure easy data access. Additionally, the algorithm can also be run on a connected lab information system. Such an algorithm is valuable when access to pathogen testing is limited and for predicting disease state in unsuspected individuals thus prompting a confirmatory test. Various pandemic and endemic applications for this algorithm are demonstrated. Methods. CBC/Diff data for pathogen positives and negatives are obtained from a preselected cohort.
• the final ML model is an ensemble of decision tree-based methods, with excellent accuracy and suitable AUROC on the test set. Significant features include eosinophil and lymphocyte counts are reported earlier. The model is able to predict hospitalized patients more accurately than non-hospitalized patients. Conclusion.
• a predictive algorithm, method, and system is provided for rapid pathogenic disease flagging using data from routine CBC testing. The algorithm is useful in pandemic or endemic scenarios as an alternate to testing, reducing testing burden and flagging unsuspected positives.
• Example 3 Prophetic example [0130] Hematology tests such as CBC/Diff (Complete Blood Count with Differential) are ubiquitously performed for overall health monitoring.
• CBC/Diff Complete Blood Count with Differential
• the present disclosure combines CBC/Diff results from the Siemens Healthineers ADVIA® 2120i Hematology System with a multiparametric machine learning algorithm for rapid flagging of sepsis positivity. Diagnostic data generation and running the algorithm on the data occur on the same instrument, to ensure easy data access. Additionally, the algorithm can also be run on a connected lab information system. Such an algorithm is valuable when access to sepsis testing is limited and for predicting sepsis state in unsuspected individuals thus prompting a confirmatory test. Various pandemic and endemic applications for this algorithm are demonstrated.
• CBC/Diff data for pathogen positives and negatives are obtained from a preselected cohort. Using data obtained therefrom, machine learning classification models are trained with 5-fold cross-validation for distinguishing positives from negatives.
• Results The final ML model is an ensemble of decision tree-based methods, with excellent accuracy and suitable AUROC on the test set. Significant features include eosinophil and lymphocyte counts are reported earlier. The model is able to predict hospitalized patients more accurately than non-hospitalized patients.
• Conclusion A predictive algorithm, method, and system is provided for rapid sepsis flagging using data from routine CBC testing.

Landscapes

• Health & Medical Sciences (AREA)
• Engineering & Computer Science (AREA)
• Public Health (AREA)
• Medical Informatics (AREA)
• General Health & Medical Sciences (AREA)
• Primary Health Care (AREA)
• Epidemiology (AREA)
• Biomedical Technology (AREA)
• Data Mining & Analysis (AREA)
• Databases & Information Systems (AREA)
• Pathology (AREA)
• Investigating Or Analysing Biological Materials (AREA)
• Medical Treatment And Welfare Office Work (AREA)

Applications Claiming Priority (4)

Publications (2)

Family

ID=87766727

Family Applications (1)

Country Status (5)

Family Cites Families (9)

• 2023
  • 2023-02-23 WO PCT/US2023/063091 patent/WO2023164525A1/en not_active Ceased
  • 2023-02-23 US US18/837,457 patent/US20250046472A1/en active Pending
  • 2023-02-23 EP EP23760903.7A patent/EP4466718A4/de active Pending
  • 2023-02-23 CN CN202380023431.7A patent/CN119137681A/zh active Pending
  • 2023-02-23 JP JP2024550311A patent/JP2025508857A/ja active Pending

Also Published As

Similar Documents

Legal Events